Customer Reviews

Causal Models: How People Think about the World and Its Alternatives

5 star:		(2)
4 star:		(1)
3 star:		(0)
2 star:		(0)
1 star:		(0)

 

 

The philosophical debate on the notion of causality has never been too much of a concern for scientists, particularly physicists who take a pragmatic attitude about cause and effect, and therefore do not get mired in the huge (and frequently useless) conceptual spaces constructed by philosophers and their apologists. The exception to this has been in some areas of theoretical physics, such as quantum mechanics and the physics of collapsed stars (black holes). In general though, it is probably fair to say that the scientific community has not been shaken by the arguments of philosophers such as David Hume, who supposedly have "demolished" some of the ideas on causality that are taken for granted by pre-Hume philosophers and the "general public."

The debate on how humans conceptualize causality and how they integrate their models of causality into decision-making however is of great interest to the scientific community, particularly psychologists and cognitive neuroscientists, who especially in the last two decades, have engaged in intensive research on this topic. A study of this research reveals that there is still a lot more to be done in this area, but what has been accomplished is impressive and fascinating. Those working in the field of artificial intelligence have taken some of these results and tried to integrate them into intelligent machines, with varying degrees of success.

For the most part, the author of this book has eschewed philosophical musings and has given the reader a view of conceptual models that is scientific and is currently in vogue in applied mathematics. Indeed, within its covers the reader will find discussions of possible worlds logic, Bayesian data modeling, and other techniques that are formulated in a framework that goes beyond the one developed in the 18th century (to paraphrase the author). The author is not shy about confronting some of the nagging issues behind how humans think about causality, but successfully avoids the trap of endless philosophical debate on the topic. Ironically though, his analysis draws on the work of some highly regarded philosophers, such as Peter Spirtes, Clark Glymour, and Richard Scheine. These philosophers have given excellent discussions of what are now called Bayesian belief networks, which have myriads of practical applications in areas such as financial and network modeling.

At least for this reviewer, the most interesting part of the book is how humans make decisions based on the causal models they develop, which as the author reminds the reader are usually based on qualitative evidence, frequently in error and fail to assess probabilities accurately (sometimes collectively called "cognitive bias"). This discussion is valuable for those readers who are actively involved in modeling real systems, both in applied and academic contexts. It sheds light for example on why managers of modeling groups insist on some sort of nontrivial time duration for the model execution, believing that to be viable a model must take an appreciable amount of time to complete in order to produce valid results. For those readers involved in models deploying discrete event simulation, it sheds light on why causal mechanisms are frequently imputed to these models, even though none can ever be found (these types of models avoid causal explanations by exhausting the realm of possibilities for the behavior of the modeled system using hypothetical randomized paths that the system may actually realize).

This book consists of two parts: Part 1 is an introduction to concepts and terminology of causal models. Part 2 consists of chapters that apply the concepts to various domains of everyday life. All chapters are written from the point of view of a computer scientist with a strong interest in philosophical foundations and social sciences (psychology, cognition and law).

The author admits that his book contains nothing new, but is an introduction for those who don't feel well prepared for advanced literature like the well-known book Causality: Models, Reasoning and Inference by Judea Pearl. Such philosophical books are often hard to read, not because of the amount of facts to remember but because they carefully examine common terms like causality to such an extent that the reader may get puzzled. The author carefully addresses the basic terminology, concepts and ideas so that the reader will be prepared for reading other books.

What I liked about the book was the detailed explanation of such fundamental concepts as conditional probability and Bayes' rule. These are not as trivial as they might look when you see them as just one formula. I also liked the explanation of Causal models, Bayesian nets and the modern theory of intervention. I was also surprised to see how much of this is closely related to the philosophical basis of law.

Why do I rate this book at 4 stars (and not 5) ? I expected more of an undergraduate textbook with a more systematic introduction. As it is, the book is good for reading in the morning while driving to work on a bus, but not as a preparation for any courses. Other authors invest much more care into writing introductory textbooks (for example Consciousness: An Introduction) and reach an even higher level of entertainment while teaching along the lines of curricula.

If you are at all interested in how we think, learn and interpret the world then read this book. It's a weighty subject, but the author brings about an ease of understanding. It'll still require some thought on your part, but it's well worth it. Very well worth the time and effort.